Hello, sorry for the late reply.
On 2/14/22 15:33, Robert Haas wrote:
On Mon, Feb 7, 2022 at 5:26 AM Frédéric Yhuel <frederic.yh...@dalibo.com> wrote:
I noticed that referential integrity checks aren't currently
parallelized. Is it on purpose?
It's not 100% clear to me that it is safe. But on the other hand, it's
also not 100% clear to me that it is unsafe.
Generally, I only added CURSOR_OPT_PARALLEL_OK in places where I was
confident that nothing bad would happen, and this wasn't one of those
places. It's something of a nested-query environment -- your criterion
#6. How do we know that the surrounding query isn't already parallel?
Perhaps because it must be DML,
Did you mean DDL?
but I think it must be more important
to support parallel DML than to support this.
I'm not sure what the right thing to do here is, but I think it would
be good if your patch included a test case.
Would that be a regression test? (in src/test/regress ?)
If yes, what should I check? Is it good enough to load auto_explain and
check that the query triggered by a foreign key addition is parallelized?
Best regards,
Frédéric
